Convolutional Neural Networks

Project: Write an Algorithm for a Dog Identification App


In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.


Why We're Here

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

Sample Dog Output

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

The Road Ahead

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

  • Step 0: Import Datasets
  • Step 1: Detect Humans
  • Step 2: Detect Dogs
  • Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
  • Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 6: Write your Algorithm
  • Step 7: Test Your Algorithm

Step 0: Import Datasets

Import Dog Dataset

In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:

  • train_files, valid_files, test_files - numpy arrays containing file paths to images
  • train_targets, valid_targets, test_targets - numpy arrays containing onehot-encoded classification labels
  • dog_names - list of string-valued dog breed names for translating labels
In [1]:
from sklearn.datasets import load_files       
from keras.utils import np_utils
import numpy as np
from glob import glob

# define function to load train, test, and validation datasets
def load_dataset(path):
    data = load_files(path)
    dog_files = np.array(data['filenames'])
    dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
    return dog_files, dog_targets

# load train, test, and validation datasets
train_files, train_targets = load_dataset('/data/dog_images/train')
valid_files, valid_targets = load_dataset('/data/dog_images/valid')
test_files, test_targets = load_dataset('/data/dog_images/test')

# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("/data/dog_images/train/*/"))]

# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))
Using TensorFlow backend.
There are 133 total dog categories.
There are 8351 total dog images.

There are 6680 training dog images.
There are 835 validation dog images.
There are 836 test dog images.

Import Human Dataset

In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.

In [2]:
import random
random.seed(8675309)

# load filenames in shuffled human dataset
human_files = np.array(glob("/data/lfw/*/*"))
random.shuffle(human_files)

# print statistics about the dataset
print('There are %d total human images.' % len(human_files))
There are 13233 total human images.

Step 1: Detect Humans

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.

In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [3]:
import cv2                
import matplotlib.pyplot as plt                        
%matplotlib inline                               

# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')

# load color (BGR) image
img = cv2.imread(human_files[3])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image
faces = face_cascade.detectMultiScale(gray)

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
    # add bounding box to color image
    cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Number of faces detected: 1

Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Write a Human Face Detector

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [4]:
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray)
    return len(faces) > 0

(IMPLEMENTATION) Assess the Human Face Detector

Question 1: Use the code cell below to test the performance of the face_detector function.

  • What percentage of the first 100 images in human_files have a detected human face?
  • What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

Answer: From below code, we can see that out of 100 images in 'human_files', 100% of the images had a human face detected. 11% of the first 100 images in 'dog_files' had a human face detected

In [5]:
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.

## TODO: Test the performance of the face_detector algorithm 
## on the images in human_files_short and dog_files_short.
h=0
for p in human_files_short:
    if face_detector(p):
        h += 1
print("human files performance : {}%".format(h))
d=0
for p in dog_files_short:
    if face_detector(p):
        d += 1
print("Dog files performance : {}%".format(d))
human files performance : 100%
Dog files performance : 11%

Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?

Answer: No it is not a reasonable expectation to have humans clearly present their face in a photo for it to be detected. This assumption doesn't seem fair to me as it adds constraints to the system. The alternative is to use deep learning to extract features that distinguish humans from dogs even if full face is not presented. Rather we should build our algorithm based on CNN with a training data that should include a diverse set of images from a wide variety of angles, lighting conditions, and partial obscurations.

In my opinion, I think we can use Haar cascades for detecting human faces. A Haar Cascade is basically a classifier which is used to detect the object for which it has been trained for, from the source. The Haar Cascade is trained by superimposing the positive image over a set of negative images. It is well known for being able to detect faces and body parts in an image, but can be trained to identify almost any object. Better results are obtained by using high quality images and increasing the amount of stages for which the classifier is trained.

We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.

In [ ]:
## (Optional) TODO: Report the performance of another  
## face detection algorithm on the LFW dataset
### Feel free to use as many code cells as needed.

Step 2: Detect Dogs

In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.

In [6]:
from keras.applications.resnet50 import ResNet50

# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')
Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels.h5
102858752/102853048 [==============================] - 2s 0us/step

Pre-process the Data

When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape

$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$

where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.

The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape

$$ (1, 224, 224, 3). $$

The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape

$$ (\text{nb_samples}, 224, 224, 3). $$

Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!

In [7]:
from keras.preprocessing import image                  
from tqdm import tqdm

def path_to_tensor(img_path):
    # loads RGB image as PIL.Image.Image type
    img = image.load_img(img_path, target_size=(224, 224))
    # convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
    x = image.img_to_array(img)
    # convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
    return np.expand_dims(x, axis=0)

def paths_to_tensor(img_paths):
    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
    return np.vstack(list_of_tensors)

Making Predictions with ResNet-50

Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.

Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.

By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.

In [8]:
from keras.applications.resnet50 import preprocess_input, decode_predictions

def ResNet50_predict_labels(img_path):
    # returns prediction vector for image located at img_path
    img = preprocess_input(path_to_tensor(img_path))
    return np.argmax(ResNet50_model.predict(img))

Write a Dog Detector

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).

We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [9]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
    prediction = ResNet50_predict_labels(img_path)
    return ((prediction <= 268) & (prediction >= 151)) 

(IMPLEMENTATION) Assess the Dog Detector

Question 3: Use the code cell below to test the performance of your dog_detector function.

  • What percentage of the images in human_files_short have a detected dog?
  • What percentage of the images in dog_files_short have a detected dog?

Answer: Percentage of images in human_files_short have a detected dog : 0%.

Percentage of images in dog_files_short have a detected dog : 100%

In [10]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
h=0
for p in human_files_short:
    if dog_detector(p):
        h += 1
print("Percentage of images in human_files_short have a detected dog : {}%".format(h))
d=0
for p in dog_files_short:
    if dog_detector(p):
        d += 1
print("Percentage of images in dog_files_short have a detected dog : {}%".format(d))
Percentage of images in human_files_short have a detected dog : 0%
Percentage of images in dog_files_short have a detected dog : 100%

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

Yellow Labrador Chocolate Labrador Black Labrador

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

Pre-process the Data

We rescale the images by dividing every pixel in every image by 255.

In [11]:
from PIL import ImageFile                            
ImageFile.LOAD_TRUNCATED_IMAGES = True                 

# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255
100%|██████████| 6680/6680 [01:14<00:00, 89.65it/s] 
100%|██████████| 835/835 [00:08<00:00, 99.45it/s] 
100%|██████████| 836/836 [00:08<00:00, 100.30it/s]

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    model.summary()

We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:

Sample CNN

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.

Answer:

  1. 3 convolutional layers are created with 3 max pooling layers in between them to learn hierarchy of high level features. Max pooling layer is added to reduce the dimensionality
  2. Flatten layer is added to reduce the matrix to row vector. This is because fully connected layer only accepts row vector.
  3. Filters were used 16, 32, 64 in each of the convolutional layers.
  4. Dropout was used along with flattening layer before using the fully connected layer to reduce overfitting and ensure that the network generalizes well.
  5. Number of noders in the last fully connected layer were setup as 133 along with softmax activation function to obtain probabilities of the prediction.
  6. Relu activation function was used for all other layers.
In [12]:
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential

model = Sequential()

### TODO: Define your architecture.
model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))

model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))

model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))

model.add(Dropout(0.4))
model.add(Flatten())
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.4))
model.add(Dense(133, activation='softmax'))
model.summary()
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 224, 224, 16)      208       
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 112, 112, 16)      0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 112, 112, 32)      2080      
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 56, 56, 32)        0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 56, 56, 64)        8256      
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 28, 28, 64)        0         
_________________________________________________________________
dropout_1 (Dropout)          (None, 28, 28, 64)        0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 50176)             0         
_________________________________________________________________
dense_1 (Dense)              (None, 512)               25690624  
_________________________________________________________________
dropout_2 (Dropout)          (None, 512)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 133)               68229     
=================================================================
Total params: 25,769,397
Trainable params: 25,769,397
Non-trainable params: 0
_________________________________________________________________

Compile the Model

In [13]:
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [14]:
from keras.callbacks import ModelCheckpoint  

### TODO: specify the number of epochs that you would like to use to train the model.

epochs = 25

### Do NOT modify the code below this line.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5', 
                               verbose=1, save_best_only=True)

model.fit(train_tensors, train_targets, 
          validation_data=(valid_tensors, valid_targets),
          epochs=epochs, batch_size=20, callbacks=[checkpointer], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/25
6660/6680 [============================>.] - ETA: 0s - loss: 4.8532 - acc: 0.0188Epoch 00001: val_loss improved from inf to 4.57251, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 36s 5ms/step - loss: 4.8524 - acc: 0.0189 - val_loss: 4.5725 - val_acc: 0.0467
Epoch 2/25
6660/6680 [============================>.] - ETA: 0s - loss: 4.3874 - acc: 0.0542Epoch 00002: val_loss improved from 4.57251 to 4.26903, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 35s 5ms/step - loss: 4.3881 - acc: 0.0542 - val_loss: 4.2690 - val_acc: 0.0659
Epoch 3/25
6660/6680 [============================>.] - ETA: 0s - loss: 3.9779 - acc: 0.1105Epoch 00003: val_loss improved from 4.26903 to 4.22114, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 35s 5ms/step - loss: 3.9776 - acc: 0.1105 - val_loss: 4.2211 - val_acc: 0.0862
Epoch 4/25
6660/6680 [============================>.] - ETA: 0s - loss: 3.3638 - acc: 0.2180- ETA: 0s - loss: 3.3667 - acc: 0Epoch 00004: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 3.3616 - acc: 0.2183 - val_loss: 4.2408 - val_acc: 0.1102
Epoch 5/25
6660/6680 [============================>.] - ETA: 0s - loss: 2.5431 - acc: 0.3856Epoch 00005: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 2.5431 - acc: 0.3855 - val_loss: 4.4959 - val_acc: 0.1006
Epoch 6/25
6660/6680 [============================>.] - ETA: 0s - loss: 1.6925 - acc: 0.5694Epoch 00006: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 1.6928 - acc: 0.5693 - val_loss: 4.9842 - val_acc: 0.1078
Epoch 7/25
6660/6680 [============================>.] - ETA: 0s - loss: 1.0426 - acc: 0.7225Epoch 00007: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 1.0430 - acc: 0.7228 - val_loss: 5.9261 - val_acc: 0.1030
Epoch 8/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.6881 - acc: 0.8173Epoch 00008: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.6896 - acc: 0.8169 - val_loss: 5.9348 - val_acc: 0.0922
Epoch 9/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.4690 - acc: 0.8769Epoch 00009: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.4692 - acc: 0.8768 - val_loss: 6.5628 - val_acc: 0.1018
Epoch 10/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.3825 - acc: 0.9012Epoch 00010: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.3821 - acc: 0.9012 - val_loss: 7.5359 - val_acc: 0.1054
Epoch 11/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.3344 - acc: 0.9096Epoch 00011: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.3364 - acc: 0.9094 - val_loss: 6.9880 - val_acc: 0.0982
Epoch 12/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2896 - acc: 0.9269Epoch 00012: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2894 - acc: 0.9269 - val_loss: 7.3558 - val_acc: 0.0970
Epoch 13/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2514 - acc: 0.9377Epoch 00013: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2511 - acc: 0.9379 - val_loss: 7.9918 - val_acc: 0.1066
Epoch 14/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2467 - acc: 0.9380Epoch 00014: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2462 - acc: 0.9382 - val_loss: 7.5528 - val_acc: 0.0922
Epoch 15/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2601 - acc: 0.9393Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2596 - acc: 0.9394 - val_loss: 8.1988 - val_acc: 0.1102
Epoch 16/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2411 - acc: 0.9456Epoch 00016: val_loss did not improve
6680/6680 [==============================] - 34s 5ms/step - loss: 0.2422 - acc: 0.9457 - val_loss: 7.4076 - val_acc: 0.1006
Epoch 17/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2350 - acc: 0.9434- ETA: 0s - loss: 0.2353 - acc: 0.9Epoch 00017: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2357 - acc: 0.9433 - val_loss: 7.8203 - val_acc: 0.0826
Epoch 18/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2359 - acc: 0.9459Epoch 00018: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2356 - acc: 0.9460 - val_loss: 7.4024 - val_acc: 0.0958
Epoch 19/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2297 - acc: 0.9465Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 36s 5ms/step - loss: 0.2293 - acc: 0.9466 - val_loss: 7.5098 - val_acc: 0.0922
Epoch 20/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2274 - acc: 0.9479Epoch 00020: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2268 - acc: 0.9481 - val_loss: 8.5843 - val_acc: 0.1054
Epoch 21/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2191 - acc: 0.9486Epoch 00021: val_loss did not improve
6680/6680 [==============================] - 36s 5ms/step - loss: 0.2197 - acc: 0.9485 - val_loss: 6.8272 - val_acc: 0.0731
Epoch 22/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2306 - acc: 0.9470Epoch 00022: val_loss did not improve
6680/6680 [==============================] - 36s 5ms/step - loss: 0.2301 - acc: 0.9470 - val_loss: 9.3590 - val_acc: 0.1018
Epoch 23/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2422 - acc: 0.9470Epoch 00023: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2422 - acc: 0.9470 - val_loss: 8.4573 - val_acc: 0.1030
Epoch 24/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2612 - acc: 0.9398Epoch 00024: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2605 - acc: 0.9400 - val_loss: 9.1214 - val_acc: 0.0934
Epoch 25/25
6660/6680 [============================>.] - ETA: 0s - loss: 0.2463 - acc: 0.9468Epoch 00025: val_loss did not improve
6680/6680 [==============================] - 35s 5ms/step - loss: 0.2465 - acc: 0.9467 - val_loss: 8.3521 - val_acc: 0.0934
Out[14]:
<keras.callbacks.History at 0x7faba9a454e0>

Load the Model with the Best Validation Loss

In [15]:
model.load_weights('saved_models/weights.best.from_scratch.hdf5')

Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.

In [16]:
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]

# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 7.8947%

Step 4: Use a CNN to Classify Dog Breeds

To reduce training time without sacrificing accuracy, we show you how to train a CNN using transfer learning. In the following step, you will get a chance to use transfer learning to train your own CNN.

Obtain Bottleneck Features

In [17]:
bottleneck_features = np.load('/data/bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']

Model Architecture

The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.

In [18]:
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))

VGG16_model.summary()
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_1 ( (None, 512)               0         
_________________________________________________________________
dense_3 (Dense)              (None, 133)               68229     
=================================================================
Total params: 68,229
Trainable params: 68,229
Non-trainable params: 0
_________________________________________________________________

Compile the Model

In [19]:
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

Train the Model

In [20]:
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5', 
                               verbose=1, save_best_only=True)

VGG16_model.fit(train_VGG16, train_targets, 
          validation_data=(valid_VGG16, valid_targets),
          epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/20
6500/6680 [============================>.] - ETA: 0s - loss: 12.9960 - acc: 0.1042Epoch 00001: val_loss improved from inf to 11.80119, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 370us/step - loss: 12.9708 - acc: 0.1060 - val_loss: 11.8012 - val_acc: 0.1629
Epoch 2/20
6500/6680 [============================>.] - ETA: 0s - loss: 11.1311 - acc: 0.2385Epoch 00002: val_loss improved from 11.80119 to 11.09410, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 322us/step - loss: 11.1544 - acc: 0.2380 - val_loss: 11.0941 - val_acc: 0.2455
Epoch 3/20
6560/6680 [============================>.] - ETA: 0s - loss: 10.5827 - acc: 0.2950Epoch 00003: val_loss improved from 11.09410 to 10.82757, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 332us/step - loss: 10.6039 - acc: 0.2940 - val_loss: 10.8276 - val_acc: 0.2695
Epoch 4/20
6500/6680 [============================>.] - ETA: 0s - loss: 10.4100 - acc: 0.3232Epoch 00004: val_loss improved from 10.82757 to 10.68214, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 314us/step - loss: 10.3964 - acc: 0.3232 - val_loss: 10.6821 - val_acc: 0.2695
Epoch 5/20
6520/6680 [============================>.] - ETA: 0s - loss: 10.2191 - acc: 0.3319Epoch 00005: val_loss improved from 10.68214 to 10.48756, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 313us/step - loss: 10.2055 - acc: 0.3328 - val_loss: 10.4876 - val_acc: 0.2850
Epoch 6/20
6620/6680 [============================>.] - ETA: 0s - loss: 10.0610 - acc: 0.3494Epoch 00006: val_loss improved from 10.48756 to 10.47316, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 311us/step - loss: 10.0626 - acc: 0.3493 - val_loss: 10.4732 - val_acc: 0.2910
Epoch 7/20
6520/6680 [============================>.] - ETA: 0s - loss: 9.9336 - acc: 0.3653Epoch 00007: val_loss improved from 10.47316 to 10.45413, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 306us/step - loss: 9.9300 - acc: 0.3653 - val_loss: 10.4541 - val_acc: 0.2958
Epoch 8/20
6580/6680 [============================>.] - ETA: 0s - loss: 9.9276 - acc: 0.3687Epoch 00008: val_loss improved from 10.45413 to 10.39637, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 303us/step - loss: 9.9030 - acc: 0.3702 - val_loss: 10.3964 - val_acc: 0.3018
Epoch 9/20
6600/6680 [============================>.] - ETA: 0s - loss: 9.8466 - acc: 0.3759Epoch 00009: val_loss improved from 10.39637 to 10.36349, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 305us/step - loss: 9.8626 - acc: 0.3749 - val_loss: 10.3635 - val_acc: 0.3054
Epoch 10/20
6640/6680 [============================>.] - ETA: 0s - loss: 9.8053 - acc: 0.3804Epoch 00010: val_loss improved from 10.36349 to 10.30758, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 305us/step - loss: 9.8094 - acc: 0.3802 - val_loss: 10.3076 - val_acc: 0.3078
Epoch 11/20
6500/6680 [============================>.] - ETA: 0s - loss: 9.7532 - acc: 0.3845Epoch 00011: val_loss improved from 10.30758 to 10.28176, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 303us/step - loss: 9.7542 - acc: 0.3843 - val_loss: 10.2818 - val_acc: 0.3018
Epoch 12/20
6520/6680 [============================>.] - ETA: 0s - loss: 9.5714 - acc: 0.3883Epoch 00012: val_loss improved from 10.28176 to 10.05203, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 310us/step - loss: 9.5838 - acc: 0.3877 - val_loss: 10.0520 - val_acc: 0.3138
Epoch 13/20
6620/6680 [============================>.] - ETA: 0s - loss: 9.3494 - acc: 0.3991Epoch 00013: val_loss improved from 10.05203 to 9.82160, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 309us/step - loss: 9.3530 - acc: 0.3988 - val_loss: 9.8216 - val_acc: 0.3174
Epoch 14/20
6580/6680 [============================>.] - ETA: 0s - loss: 9.1168 - acc: 0.4152Epoch 00014: val_loss improved from 9.82160 to 9.65940, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 307us/step - loss: 9.1167 - acc: 0.4150 - val_loss: 9.6594 - val_acc: 0.3377
Epoch 15/20
6600/6680 [============================>.] - ETA: 0s - loss: 8.9459 - acc: 0.4297Epoch 00015: val_loss improved from 9.65940 to 9.57854, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 305us/step - loss: 8.9530 - acc: 0.4293 - val_loss: 9.5785 - val_acc: 0.3413
Epoch 16/20
6640/6680 [============================>.] - ETA: 0s - loss: 8.8439 - acc: 0.4377Epoch 00016: val_loss improved from 9.57854 to 9.50308, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 309us/step - loss: 8.8587 - acc: 0.4367 - val_loss: 9.5031 - val_acc: 0.3473
Epoch 17/20
6540/6680 [============================>.] - ETA: 0s - loss: 8.7917 - acc: 0.4456Epoch 00017: val_loss improved from 9.50308 to 9.46699, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 313us/step - loss: 8.8006 - acc: 0.4452 - val_loss: 9.4670 - val_acc: 0.3593
Epoch 18/20
6540/6680 [============================>.] - ETA: 0s - loss: 8.7794 - acc: 0.4472Epoch 00018: val_loss improved from 9.46699 to 9.38234, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 306us/step - loss: 8.7860 - acc: 0.4470 - val_loss: 9.3823 - val_acc: 0.3653
Epoch 19/20
6540/6680 [============================>.] - ETA: 0s - loss: 8.7962 - acc: 0.4506Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 2s 308us/step - loss: 8.7740 - acc: 0.4519 - val_loss: 9.4141 - val_acc: 0.3605
Epoch 20/20
6620/6680 [============================>.] - ETA: 0s - loss: 8.6659 - acc: 0.4524Epoch 00020: val_loss improved from 9.38234 to 9.30085, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 305us/step - loss: 8.6677 - acc: 0.4524 - val_loss: 9.3008 - val_acc: 0.3749
Out[20]:
<keras.callbacks.History at 0x7faba9d28668>

Load the Model with the Best Validation Loss

In [21]:
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')

Test the Model

Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.

In [22]:
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]

# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 38.3971%

Predict Dog Breed with the Model

In [23]:
from extract_bottleneck_features import *

def VGG16_predict_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
    # obtain predicted vector
    predicted_vector = VGG16_model.predict(bottleneck_feature)
    # return dog breed that is predicted by the model
    return dog_names[np.argmax(predicted_vector)]

Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras. These are already in the workspace, at /data/bottleneck_features. If you wish to download them on a different machine, they can be found at:

The files are encoded as such:

Dog{network}Data.npz

where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception.

The above architectures are downloaded and stored for you in the /data/bottleneck_features/ folder.

This means the following will be in the /data/bottleneck_features/ folder:

DogVGG19Data.npz DogResnet50Data.npz DogInceptionV3Data.npz DogXceptionData.npz

(IMPLEMENTATION) Obtain Bottleneck Features

In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:

bottleneck_features = np.load('/data/bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']
In [24]:
### TODO: Obtain bottleneck features from another pre-trained CNN.
bottleneck_features = np.load('/data/bottleneck_features/DogXceptionData.npz')
train_Xception = bottleneck_features['train']
valid_Xception = bottleneck_features['valid']
test_Xception = bottleneck_features['test']

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    <your model's name>.summary()

Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer: The new data set is small and similar to the original training data, hence end of the network is sliced off and a fully connected layer that matches the number of classes in the new data set is added. Next the weights of the new fully connected layer are randomized; All the weights from the pre-trained network are frozen. Finally network is trained to update the weights of the new fully connected layer. Overall, I have used the Xception model which gave the best result (85.17% accuracy and 84.2 with ResNet-50). I also added a fully connected layer with 500 nodes and a ReLU activation function to detect more patterns and a Dropout to avoid overfitting.

In [25]:
### TODO: Define your architecture.
from keras.layers import Dropout

Xception_model = Sequential()
Xception_model.add(GlobalAveragePooling2D(input_shape=train_Xception.shape[1:]))
Xception_model.add(Dense(500, activation='relu'))
Xception_model.add(Dropout(0.4))
Xception_model.add(Dense(133, activation='softmax'))

Xception_model.summary()
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_2 ( (None, 2048)              0         
_________________________________________________________________
dense_4 (Dense)              (None, 500)               1024500   
_________________________________________________________________
dropout_3 (Dropout)          (None, 500)               0         
_________________________________________________________________
dense_5 (Dense)              (None, 133)               66633     
=================================================================
Total params: 1,091,133
Trainable params: 1,091,133
Non-trainable params: 0
_________________________________________________________________

(IMPLEMENTATION) Compile the Model

In [26]:
### TODO: Compile the model.
Xception_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [27]:
### TODO: Train the model.
# from keras.preprocessing.image import ImageDataGenerator
# datagen = ImageDataGenerator(width_shift_range=0.2,height_shift_range=0.2,horizontal_flip=True)
# datagen.fit(train_Resnet50)

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.Xception.hdf5', 
                               verbose=1, save_best_only=True)

Xception_model.fit(train_Xception, train_targets, 
          validation_data=(valid_Xception, valid_targets),
          epochs=100, batch_size=20, callbacks=[checkpointer], verbose=1)
Train on 6680 samples, validate on 835 samples
Epoch 1/100
6620/6680 [============================>.] - ETA: 0s - loss: 1.4307 - acc: 0.6489Epoch 00001: val_loss improved from inf to 0.61245, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 4s 658us/step - loss: 1.4273 - acc: 0.6499 - val_loss: 0.6125 - val_acc: 0.8156
Epoch 2/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.6565 - acc: 0.8071Epoch 00002: val_loss improved from 0.61245 to 0.60841, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 4s 614us/step - loss: 0.6553 - acc: 0.8072 - val_loss: 0.6084 - val_acc: 0.8240
Epoch 3/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.5300 - acc: 0.8376Epoch 00003: val_loss did not improve
6680/6680 [==============================] - 4s 604us/step - loss: 0.5295 - acc: 0.8371 - val_loss: 0.6400 - val_acc: 0.8084
Epoch 4/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.4458 - acc: 0.8660Epoch 00004: val_loss did not improve
6680/6680 [==============================] - 4s 599us/step - loss: 0.4457 - acc: 0.8659 - val_loss: 0.6235 - val_acc: 0.8240
Epoch 5/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.3704 - acc: 0.8840Epoch 00005: val_loss improved from 0.60841 to 0.59717, saving model to saved_models/weights.best.Xception.hdf5
6680/6680 [==============================] - 4s 605us/step - loss: 0.3741 - acc: 0.8834 - val_loss: 0.5972 - val_acc: 0.8479
Epoch 6/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.3453 - acc: 0.8971Epoch 00006: val_loss did not improve
6680/6680 [==============================] - 4s 605us/step - loss: 0.3493 - acc: 0.8964 - val_loss: 0.6005 - val_acc: 0.8383
Epoch 7/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.3099 - acc: 0.9088Epoch 00007: val_loss did not improve
6680/6680 [==============================] - 4s 602us/step - loss: 0.3100 - acc: 0.9087 - val_loss: 0.6733 - val_acc: 0.8275
Epoch 8/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.2714 - acc: 0.9159Epoch 00008: val_loss did not improve
6680/6680 [==============================] - 4s 603us/step - loss: 0.2704 - acc: 0.9157 - val_loss: 0.7746 - val_acc: 0.8263
Epoch 9/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.2664 - acc: 0.9178Epoch 00009: val_loss did not improve
6680/6680 [==============================] - 4s 601us/step - loss: 0.2673 - acc: 0.9177 - val_loss: 0.7632 - val_acc: 0.8371
Epoch 10/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.2412 - acc: 0.9282Epoch 00010: val_loss did not improve
6680/6680 [==============================] - 4s 601us/step - loss: 0.2393 - acc: 0.9289 - val_loss: 0.8245 - val_acc: 0.8395
Epoch 11/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.2110 - acc: 0.9367Epoch 00011: val_loss did not improve
6680/6680 [==============================] - 4s 589us/step - loss: 0.2106 - acc: 0.9368 - val_loss: 0.8185 - val_acc: 0.8527
Epoch 12/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.2071 - acc: 0.9378Epoch 00012: val_loss did not improve
6680/6680 [==============================] - 4s 582us/step - loss: 0.2074 - acc: 0.9379 - val_loss: 0.8628 - val_acc: 0.8407
Epoch 13/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1945 - acc: 0.9400Epoch 00013: val_loss did not improve
6680/6680 [==============================] - 4s 608us/step - loss: 0.1954 - acc: 0.9400 - val_loss: 0.8411 - val_acc: 0.8323
Epoch 14/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1845 - acc: 0.9458Epoch 00014: val_loss did not improve
6680/6680 [==============================] - 4s 616us/step - loss: 0.1848 - acc: 0.9460 - val_loss: 0.9509 - val_acc: 0.8311
Epoch 15/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1705 - acc: 0.9505Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 4s 618us/step - loss: 0.1715 - acc: 0.9503 - val_loss: 0.9133 - val_acc: 0.8491
Epoch 16/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1576 - acc: 0.9506Epoch 00016: val_loss did not improve
6680/6680 [==============================] - 4s 598us/step - loss: 0.1602 - acc: 0.9501 - val_loss: 0.9861 - val_acc: 0.8407
Epoch 17/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1493 - acc: 0.9585Epoch 00017: val_loss did not improve
6680/6680 [==============================] - 4s 612us/step - loss: 0.1491 - acc: 0.9582 - val_loss: 0.9361 - val_acc: 0.8515
Epoch 18/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1447 - acc: 0.9591Epoch 00018: val_loss did not improve
6680/6680 [==============================] - 4s 607us/step - loss: 0.1458 - acc: 0.9587 - val_loss: 0.9937 - val_acc: 0.8347
Epoch 19/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1350 - acc: 0.9609Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 4s 603us/step - loss: 0.1342 - acc: 0.9611 - val_loss: 1.0115 - val_acc: 0.8443
Epoch 20/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1460 - acc: 0.9609Epoch 00020: val_loss did not improve
6680/6680 [==============================] - 4s 587us/step - loss: 0.1456 - acc: 0.9611 - val_loss: 1.0585 - val_acc: 0.8359
Epoch 21/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1221 - acc: 0.9634Epoch 00021: val_loss did not improve
6680/6680 [==============================] - 4s 594us/step - loss: 0.1245 - acc: 0.9629 - val_loss: 1.1191 - val_acc: 0.8395
Epoch 22/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.1270 - acc: 0.9636Epoch 00022: val_loss did not improve
6680/6680 [==============================] - 4s 611us/step - loss: 0.1263 - acc: 0.9636 - val_loss: 1.1467 - val_acc: 0.8479
Epoch 23/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1275 - acc: 0.9674Epoch 00023: val_loss did not improve
6680/6680 [==============================] - 4s 623us/step - loss: 0.1285 - acc: 0.9671 - val_loss: 1.1690 - val_acc: 0.8407
Epoch 24/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1268 - acc: 0.9650Epoch 00024: val_loss did not improve
6680/6680 [==============================] - 4s 622us/step - loss: 0.1267 - acc: 0.9650 - val_loss: 1.1130 - val_acc: 0.8467
Epoch 25/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1184 - acc: 0.9687Epoch 00025: val_loss did not improve
6680/6680 [==============================] - 4s 598us/step - loss: 0.1190 - acc: 0.9686 - val_loss: 1.0974 - val_acc: 0.8431
Epoch 26/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1247 - acc: 0.9677Epoch 00026: val_loss did not improve
6680/6680 [==============================] - 4s 599us/step - loss: 0.1242 - acc: 0.9677 - val_loss: 1.1071 - val_acc: 0.8503
Epoch 27/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.1090 - acc: 0.9695Epoch 00027: val_loss did not improve
6680/6680 [==============================] - 4s 586us/step - loss: 0.1078 - acc: 0.9699 - val_loss: 1.1976 - val_acc: 0.8335
Epoch 28/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1100 - acc: 0.9716Epoch 00028: val_loss did not improve
6680/6680 [==============================] - 4s 597us/step - loss: 0.1105 - acc: 0.9717 - val_loss: 1.2492 - val_acc: 0.8431
Epoch 29/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1064 - acc: 0.9719Epoch 00029: val_loss did not improve
6680/6680 [==============================] - 4s 616us/step - loss: 0.1057 - acc: 0.9720 - val_loss: 1.2575 - val_acc: 0.8371
Epoch 30/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1030 - acc: 0.9728Epoch 00030: val_loss did not improve
6680/6680 [==============================] - 4s 613us/step - loss: 0.1031 - acc: 0.9728 - val_loss: 1.2678 - val_acc: 0.8311
Epoch 31/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0886 - acc: 0.9755Epoch 00031: val_loss did not improve
6680/6680 [==============================] - 4s 606us/step - loss: 0.0879 - acc: 0.9756 - val_loss: 1.2094 - val_acc: 0.8407
Epoch 32/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0953 - acc: 0.9711Epoch 00032: val_loss did not improve
6680/6680 [==============================] - 4s 604us/step - loss: 0.0955 - acc: 0.9710 - val_loss: 1.1937 - val_acc: 0.8467
Epoch 33/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0888 - acc: 0.9767Epoch 00033: val_loss did not improve
6680/6680 [==============================] - 4s 607us/step - loss: 0.0880 - acc: 0.9769 - val_loss: 1.2898 - val_acc: 0.8467
Epoch 34/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0914 - acc: 0.9772Epoch 00034: val_loss did not improve
6680/6680 [==============================] - 4s 603us/step - loss: 0.0910 - acc: 0.9772 - val_loss: 1.2026 - val_acc: 0.8479
Epoch 35/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0892 - acc: 0.9769Epoch 00035: val_loss did not improve
6680/6680 [==============================] - 4s 613us/step - loss: 0.0886 - acc: 0.9769 - val_loss: 1.3201 - val_acc: 0.8515
Epoch 36/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0889 - acc: 0.9767Epoch 00036: val_loss did not improve
6680/6680 [==============================] - 4s 610us/step - loss: 0.0892 - acc: 0.9766 - val_loss: 1.3435 - val_acc: 0.8419
Epoch 37/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.1232 - acc: 0.9745Epoch 00037: val_loss did not improve
6680/6680 [==============================] - 4s 602us/step - loss: 0.1239 - acc: 0.9741 - val_loss: 1.3132 - val_acc: 0.8443
Epoch 38/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0958 - acc: 0.9763Epoch 00038: val_loss did not improve
6680/6680 [==============================] - 4s 604us/step - loss: 0.0966 - acc: 0.9762 - val_loss: 1.3275 - val_acc: 0.8467
Epoch 39/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0881 - acc: 0.9767Epoch 00039: val_loss did not improve
6680/6680 [==============================] - 4s 604us/step - loss: 0.0874 - acc: 0.9768 - val_loss: 1.2115 - val_acc: 0.8491
Epoch 40/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0828 - acc: 0.9808Epoch 00040: val_loss did not improve
6680/6680 [==============================] - 4s 607us/step - loss: 0.0823 - acc: 0.9808 - val_loss: 1.3151 - val_acc: 0.8623
Epoch 41/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0946 - acc: 0.9758Epoch 00041: val_loss did not improve
6680/6680 [==============================] - 4s 602us/step - loss: 0.0938 - acc: 0.9760 - val_loss: 1.4216 - val_acc: 0.8371
Epoch 42/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0702 - acc: 0.9813Epoch 00042: val_loss did not improve
6680/6680 [==============================] - 4s 592us/step - loss: 0.0714 - acc: 0.9813 - val_loss: 1.3126 - val_acc: 0.8539
Epoch 43/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0887 - acc: 0.9799Epoch 00043: val_loss did not improve
6680/6680 [==============================] - 4s 601us/step - loss: 0.0884 - acc: 0.9799 - val_loss: 1.4612 - val_acc: 0.8383
Epoch 44/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0796 - acc: 0.9811Epoch 00044: val_loss did not improve
6680/6680 [==============================] - 4s 597us/step - loss: 0.0792 - acc: 0.9811 - val_loss: 1.3859 - val_acc: 0.8551
Epoch 45/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0695 - acc: 0.9832Epoch 00045: val_loss did not improve
6680/6680 [==============================] - 4s 595us/step - loss: 0.0694 - acc: 0.9831 - val_loss: 1.4439 - val_acc: 0.8455
Epoch 46/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0705 - acc: 0.9819Epoch 00046: val_loss did not improve
6680/6680 [==============================] - 4s 593us/step - loss: 0.0712 - acc: 0.9816 - val_loss: 1.3731 - val_acc: 0.8503
Epoch 47/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0765 - acc: 0.9817Epoch 00047: val_loss did not improve
6680/6680 [==============================] - 4s 602us/step - loss: 0.0758 - acc: 0.9819 - val_loss: 1.3284 - val_acc: 0.8491
Epoch 48/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0749 - acc: 0.9808Epoch 00048: val_loss did not improve
6680/6680 [==============================] - 4s 598us/step - loss: 0.0745 - acc: 0.9807 - val_loss: 1.3940 - val_acc: 0.8395
Epoch 49/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0685 - acc: 0.9838Epoch 00049: val_loss did not improve
6680/6680 [==============================] - 4s 598us/step - loss: 0.0696 - acc: 0.9835 - val_loss: 1.4596 - val_acc: 0.8323
Epoch 50/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0779 - acc: 0.9823Epoch 00050: val_loss did not improve
6680/6680 [==============================] - 4s 600us/step - loss: 0.0773 - acc: 0.9825 - val_loss: 1.5794 - val_acc: 0.8323
Epoch 51/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0732 - acc: 0.9837Epoch 00051: val_loss did not improve
6680/6680 [==============================] - 4s 596us/step - loss: 0.0725 - acc: 0.9838 - val_loss: 1.5538 - val_acc: 0.8431
Epoch 52/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0641 - acc: 0.9853Epoch 00052: val_loss did not improve
6680/6680 [==============================] - 4s 600us/step - loss: 0.0636 - acc: 0.9855 - val_loss: 1.3499 - val_acc: 0.8575
Epoch 53/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0786 - acc: 0.9843Epoch 00053: val_loss did not improve
6680/6680 [==============================] - 4s 596us/step - loss: 0.0786 - acc: 0.9843 - val_loss: 1.3940 - val_acc: 0.8455
Epoch 54/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.0672 - acc: 0.9850Epoch 00054: val_loss did not improve
6680/6680 [==============================] - 4s 598us/step - loss: 0.0664 - acc: 0.9852 - val_loss: 1.3850 - val_acc: 0.8455
Epoch 55/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0731 - acc: 0.9834Epoch 00055: val_loss did not improve
6680/6680 [==============================] - 4s 595us/step - loss: 0.0736 - acc: 0.9832 - val_loss: 1.3927 - val_acc: 0.8587
Epoch 56/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0800 - acc: 0.9832Epoch 00056: val_loss did not improve
6680/6680 [==============================] - 4s 597us/step - loss: 0.0793 - acc: 0.9834 - val_loss: 1.3751 - val_acc: 0.8443
Epoch 57/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0890 - acc: 0.9807Epoch 00057: val_loss did not improve
6680/6680 [==============================] - 4s 588us/step - loss: 0.0883 - acc: 0.9808 - val_loss: 1.3796 - val_acc: 0.8407
Epoch 58/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0735 - acc: 0.9843Epoch 00058: val_loss did not improve
6680/6680 [==============================] - 4s 592us/step - loss: 0.0729 - acc: 0.9844 - val_loss: 1.4524 - val_acc: 0.8491
Epoch 59/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0890 - acc: 0.9820Epoch 00059: val_loss did not improve
6680/6680 [==============================] - 4s 578us/step - loss: 0.0891 - acc: 0.9820 - val_loss: 1.3795 - val_acc: 0.8479
Epoch 60/100
6640/6680 [============================>.] - ETA: 0s - loss: 0.0691 - acc: 0.9848Epoch 00060: val_loss did not improve
6680/6680 [==============================] - 4s 587us/step - loss: 0.0690 - acc: 0.9847 - val_loss: 1.4103 - val_acc: 0.8635
Epoch 61/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0763 - acc: 0.9841Epoch 00061: val_loss did not improve
6680/6680 [==============================] - 4s 586us/step - loss: 0.0768 - acc: 0.9841 - val_loss: 1.4617 - val_acc: 0.8407
Epoch 62/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0740 - acc: 0.9850Epoch 00062: val_loss did not improve
6680/6680 [==============================] - 4s 591us/step - loss: 0.0734 - acc: 0.9852 - val_loss: 1.5016 - val_acc: 0.8395
Epoch 63/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.0647 - acc: 0.9830Epoch 00063: val_loss did not improve
6680/6680 [==============================] - 4s 593us/step - loss: 0.0646 - acc: 0.9831 - val_loss: 1.4731 - val_acc: 0.8503
Epoch 64/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0715 - acc: 0.9841Epoch 00064: val_loss did not improve
6680/6680 [==============================] - 4s 587us/step - loss: 0.0710 - acc: 0.9841 - val_loss: 1.4267 - val_acc: 0.8539
Epoch 65/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0646 - acc: 0.9849Epoch 00065: val_loss did not improve
6680/6680 [==============================] - 4s 587us/step - loss: 0.0641 - acc: 0.9850 - val_loss: 1.4446 - val_acc: 0.8503
Epoch 66/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0669 - acc: 0.9849Epoch 00066: val_loss did not improve
6680/6680 [==============================] - 4s 585us/step - loss: 0.0667 - acc: 0.9849 - val_loss: 1.4629 - val_acc: 0.8515
Epoch 67/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0772 - acc: 0.9853Epoch 00067: val_loss did not improve
6680/6680 [==============================] - 4s 585us/step - loss: 0.0772 - acc: 0.9853 - val_loss: 1.4642 - val_acc: 0.8539
Epoch 68/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0757 - acc: 0.9855Epoch 00068: val_loss did not improve
6680/6680 [==============================] - 4s 588us/step - loss: 0.0755 - acc: 0.9853 - val_loss: 1.4844 - val_acc: 0.8575
Epoch 69/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0704 - acc: 0.9844Epoch 00069: val_loss did not improve
6680/6680 [==============================] - 4s 586us/step - loss: 0.0700 - acc: 0.9844 - val_loss: 1.4417 - val_acc: 0.8431
Epoch 70/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0673 - acc: 0.9863Epoch 00070: val_loss did not improve
6680/6680 [==============================] - 4s 590us/step - loss: 0.0670 - acc: 0.9862 - val_loss: 1.5337 - val_acc: 0.8371
Epoch 71/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0561 - acc: 0.9870Epoch 00071: val_loss did not improve
6680/6680 [==============================] - 4s 586us/step - loss: 0.0558 - acc: 0.9870 - val_loss: 1.4924 - val_acc: 0.8527
Epoch 72/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0616 - acc: 0.9867Epoch 00072: val_loss did not improve
6680/6680 [==============================] - 4s 586us/step - loss: 0.0612 - acc: 0.9868 - val_loss: 1.5098 - val_acc: 0.8515
Epoch 73/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.0704 - acc: 0.9847Epoch 00073: val_loss did not improve
6680/6680 [==============================] - 4s 597us/step - loss: 0.0696 - acc: 0.9849 - val_loss: 1.5058 - val_acc: 0.8407
Epoch 74/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0529 - acc: 0.9864Epoch 00074: val_loss did not improve
6680/6680 [==============================] - 4s 591us/step - loss: 0.0527 - acc: 0.9864 - val_loss: 1.4665 - val_acc: 0.8407
Epoch 75/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0625 - acc: 0.9867Epoch 00075: val_loss did not improve
6680/6680 [==============================] - 4s 592us/step - loss: 0.0621 - acc: 0.9868 - val_loss: 1.4348 - val_acc: 0.8455
Epoch 76/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0665 - acc: 0.9866Epoch 00076: val_loss did not improve
6680/6680 [==============================] - 4s 600us/step - loss: 0.0668 - acc: 0.9865 - val_loss: 1.4814 - val_acc: 0.8599
Epoch 77/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0573 - acc: 0.9885Epoch 00077: val_loss did not improve
6680/6680 [==============================] - 4s 591us/step - loss: 0.0587 - acc: 0.9885 - val_loss: 1.5617 - val_acc: 0.8443
Epoch 78/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0655 - acc: 0.9861Epoch 00078: val_loss did not improve
6680/6680 [==============================] - 4s 590us/step - loss: 0.0654 - acc: 0.9861 - val_loss: 1.5773 - val_acc: 0.8527
Epoch 79/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.0627 - acc: 0.9870Epoch 00079: val_loss did not improve
6680/6680 [==============================] - 4s 592us/step - loss: 0.0622 - acc: 0.9870 - val_loss: 1.5906 - val_acc: 0.8443
Epoch 80/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0549 - acc: 0.9881Epoch 00080: val_loss did not improve
6680/6680 [==============================] - 4s 588us/step - loss: 0.0583 - acc: 0.9879 - val_loss: 1.4321 - val_acc: 0.8563
Epoch 81/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0506 - acc: 0.9903Epoch 00081: val_loss did not improve
6680/6680 [==============================] - 4s 594us/step - loss: 0.0501 - acc: 0.9904 - val_loss: 1.6204 - val_acc: 0.8419
Epoch 82/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0610 - acc: 0.9867Epoch 00082: val_loss did not improve
6680/6680 [==============================] - 4s 594us/step - loss: 0.0630 - acc: 0.9867 - val_loss: 1.5428 - val_acc: 0.8503
Epoch 83/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0524 - acc: 0.9881Epoch 00083: val_loss did not improve
6680/6680 [==============================] - 4s 590us/step - loss: 0.0524 - acc: 0.9880 - val_loss: 1.6495 - val_acc: 0.8443
Epoch 84/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0539 - acc: 0.9891Epoch 00084: val_loss did not improve
6680/6680 [==============================] - 4s 589us/step - loss: 0.0536 - acc: 0.9891 - val_loss: 1.5184 - val_acc: 0.8515
Epoch 85/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0567 - acc: 0.9879Epoch 00085: val_loss did not improve
6680/6680 [==============================] - 4s 592us/step - loss: 0.0562 - acc: 0.9880 - val_loss: 1.5877 - val_acc: 0.8527
Epoch 86/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0604 - acc: 0.9885Epoch 00086: val_loss did not improve
6680/6680 [==============================] - 4s 594us/step - loss: 0.0618 - acc: 0.9883 - val_loss: 1.5620 - val_acc: 0.8515
Epoch 87/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0557 - acc: 0.9888Epoch 00087: val_loss did not improve
6680/6680 [==============================] - 4s 588us/step - loss: 0.0561 - acc: 0.9888 - val_loss: 1.5634 - val_acc: 0.8599
Epoch 88/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0456 - acc: 0.9902Epoch 00088: val_loss did not improve
6680/6680 [==============================] - 4s 576us/step - loss: 0.0460 - acc: 0.9901 - val_loss: 1.6571 - val_acc: 0.8395
Epoch 89/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0531 - acc: 0.9905Epoch 00089: val_loss did not improve
6680/6680 [==============================] - 4s 590us/step - loss: 0.0527 - acc: 0.9906 - val_loss: 1.6903 - val_acc: 0.8371
Epoch 90/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0588 - acc: 0.9879Epoch 00090: val_loss did not improve
6680/6680 [==============================] - 4s 583us/step - loss: 0.0602 - acc: 0.9876 - val_loss: 1.6502 - val_acc: 0.8587
Epoch 91/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0677 - acc: 0.9870Epoch 00091: val_loss did not improve
6680/6680 [==============================] - 4s 593us/step - loss: 0.0671 - acc: 0.9871 - val_loss: 1.5615 - val_acc: 0.8551
Epoch 92/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0488 - acc: 0.9912Epoch 00092: val_loss did not improve
6680/6680 [==============================] - 4s 593us/step - loss: 0.0484 - acc: 0.9913 - val_loss: 1.6130 - val_acc: 0.8407
Epoch 93/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0438 - acc: 0.9899Epoch 00093: val_loss did not improve
6680/6680 [==============================] - 4s 600us/step - loss: 0.0434 - acc: 0.9900 - val_loss: 1.5400 - val_acc: 0.8491
Epoch 94/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0417 - acc: 0.9911Epoch 00094: val_loss did not improve
6680/6680 [==============================] - 4s 589us/step - loss: 0.0421 - acc: 0.9907 - val_loss: 1.5859 - val_acc: 0.8491
Epoch 95/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0564 - acc: 0.9891Epoch 00095: val_loss did not improve
6680/6680 [==============================] - 4s 596us/step - loss: 0.0561 - acc: 0.9892 - val_loss: 1.7741 - val_acc: 0.8515
Epoch 96/100
6600/6680 [============================>.] - ETA: 0s - loss: 0.0601 - acc: 0.9888Epoch 00096: val_loss did not improve
6680/6680 [==============================] - 4s 599us/step - loss: 0.0595 - acc: 0.9889 - val_loss: 1.7319 - val_acc: 0.8299
Epoch 97/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0579 - acc: 0.9897Epoch 00097: val_loss did not improve
6680/6680 [==============================] - 4s 598us/step - loss: 0.0574 - acc: 0.9898 - val_loss: 1.6606 - val_acc: 0.8407
Epoch 98/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0522 - acc: 0.9885Epoch 00098: val_loss did not improve
6680/6680 [==============================] - 4s 595us/step - loss: 0.0518 - acc: 0.9886 - val_loss: 1.6340 - val_acc: 0.8503
Epoch 99/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0601 - acc: 0.9896Epoch 00099: val_loss did not improve
6680/6680 [==============================] - 4s 596us/step - loss: 0.0597 - acc: 0.9897 - val_loss: 1.5865 - val_acc: 0.8539
Epoch 100/100
6620/6680 [============================>.] - ETA: 0s - loss: 0.0412 - acc: 0.9902Epoch 00100: val_loss did not improve
6680/6680 [==============================] - 4s 601us/step - loss: 0.0409 - acc: 0.9903 - val_loss: 1.5973 - val_acc: 0.8443
Out[27]:
<keras.callbacks.History at 0x7faba9eac9b0>

(IMPLEMENTATION) Load the Model with the Best Validation Loss

In [31]:
### TODO: Load the model weights with the best validation loss.
Xception_model.load_weights('saved_models/weights.best.Xception.hdf5')

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.

In [28]:
### TODO: Calculate classification accuracy on the test dataset.
Xception_predictions = [np.argmax(Xception_model.predict(np.expand_dims(feature, axis=0))) for feature in test_Xception]
test_accuracy = 100*np.sum(np.array(Xception_predictions)==np.argmax(test_targets, axis=1))/len(Xception_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)
Test accuracy: 85.2871%

(IMPLEMENTATION) Predict Dog Breed with the Model

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.

Similar to the analogous function in Step 5, your function should have three steps:

  1. Extract the bottleneck features corresponding to the chosen CNN model.
  2. Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
  3. Use the dog_names array defined in Step 0 of this notebook to return the corresponding breed.

The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function

extract_{network}

where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.

In [29]:
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.

def Xception_predict_breed(img_path):
    bottleneck_feature = extract_Xception(path_to_tensor(img_path))    
    predicted_vector = Xception_model.predict(bottleneck_feature)       
    return dog_names[np.argmax(predicted_vector)]

Step 6: Write your Algorithm

Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,

  • if a dog is detected in the image, return the predicted breed.
  • if a human is detected in the image, return the resembling dog breed.
  • if neither is detected in the image, provide output that indicates an error.

You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.

Some sample output for our algorithm is provided below, but feel free to design your own user experience!

Sample Human Output

(IMPLEMENTATION) Write your Algorithm

In [30]:
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.
from PIL import Image
import matplotlib.pyplot as plt
%matplotlib inline
import matplotlib.image as mpimg

def my_predictor(img_path):
    # Load the image
    img= mpimg.imread(img_path)
    plt.figure(figsize=(10,10))
    plt.imshow(img)
    
    # Use dog detector and dog_breed_predictor functions to detect dogs and predict their breed
    if dog_detector(img_path) == True:
        print("It is a dog!.Dog breed is:", Xception_predict_breed(img_path))
    elif face_detector(img_path) == True:
         print("It is a human!.You resemble the dog breed:", Xception_predict_breed(img_path))
    else:
        print("An error has occured")

Step 7: Test Your Algorithm

In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?

(IMPLEMENTATION) Test Your Algorithm on Sample Images!

Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.

Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.

Answer: The output of the model is pretty good. Better than I expected. On images it has never seen below, it was able to correctly identify 5 of the 6 dogs. However there is still scope for improvement.I have identified the following points for improvement:

  1. The model prediction accuracy is not the same for all dog classes. For some classes it only has 3-4 dog images to train. Model can be made more accurate by adding more data
  2. During several runs of the algorithm, I found that the predicted dog category for one of the images changed. It would be good if model can also output the certainity it has with the prediction
  3. Different model architectures can be explored to reduce prediction time while maintaining accuracy
In [31]:
## TODO: Execute your algorithm from Step 6 on
## at least 6 images on your computer.
## Feel free to use as many code cells as needed.
for infile in sorted(glob("../test_images/*")):
     my_predictor(infile)
Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.4/xception_weights_tf_dim_ordering_tf_kernels_notop.h5
83689472/83683744 [==============================] - 1s 0us/step
It is a dog!.Dog breed is: in/006.American_eskimo_dog
It is a human!.You resemble the dog breed: in/002.Afghan_hound
It is a human!.You resemble the dog breed: in/080.Greater_swiss_mountain_dog
It is a human!.You resemble the dog breed: in/107.Norfolk_terrier
An error has occured
An error has occured
An error has occured
It is a human!.You resemble the dog breed: in/119.Petit_basset_griffon_vendeen
It is a dog!.Dog breed is: in/051.Chow_chow
It is a dog!.Dog breed is: in/115.Papillon
An error has occured
An error has occured
An error has occured
An error has occured
An error has occured
An error has occured

Please download your notebook to submit

In order to submit, please do the following:

  1. Download an HTML version of the notebook to your computer using 'File: Download as...'
  2. Click on the orange Jupyter circle on the top left of the workspace.
  3. Navigate into the dog-project folder to ensure that you are using the provided dog_images, lfw, and bottleneck_features folders; this means that those folders will not appear in the dog-project folder. If they do appear because you downloaded them, delete them.
  4. While in the dog-project folder, upload the HTML version of this notebook you just downloaded. The upload button is on the top right.
  5. Navigate back to the home folder by clicking on the two dots next to the folder icon, and then open up a terminal under the 'new' tab on the top right
  6. Zip the dog-project folder with the following command in the terminal: zip -r dog-project.zip dog-project
  7. Download the zip file by clicking on the square next to it and selecting 'download'. This will be the zip file you turn in on the next node after this workspace!